We consider the problem of optimal evasion when the pursuer is known to employ fixed gain proportional navigation. The performance index is a measure of closest approach. The analysis is done for planar motions at constant speed. The kinematics are first linearized around a nominal collision course. The dynamics of the opponents are modeled by first order systems and their accelerations may be bounded.

Three cases are studied: unconstrained optimal evasion (where the evader is not subjected to any path constraint) against a single pursuer, optimal evasion with a terminal path angle constraint for the evader and optimal evasion against more than one pursuer.

The optimal controls are shown to be 'bang - bang' with the number of switches depending on the pursuer’s navigation gain and on the particular constraints of each case.